droneposelib

Python library for Visual-Inertial Odometry (VIO). Wrapps the C++/Eigen library DronePoseLib.

Solvers available

The current list of solvers are the following:

* Measured on a laptop with an Intel(R) Core(TM) i5-6200U CPU @ 2.30GHz

** If using the use_fast_solver option

*** Undocumented

Note that all solvers assume that all three degrees of freedom for the rotation matrices are available, i.e. potential IMU drift is ignored. By using e.g. magnetometer data one can often get good estimates, if fused with accelerometer data and gyroscope data. This approach is e.g. used in the Madgwick filter. Futhermore, we have succesfully used pre-integration techniques to achieve robust and accurate performance for shorter time frames without the need of magnetometer data.

Installation

A pre-alpha release is available at PyPi, and can be installed using

    $ pip install droneposelib

You may also compile the source code, see "Development".

Examples

See the example directory for examples on how to use the solvers.

Development

You are more than welcome to contribute your our other relevant solvers. More info soon.

References

The code is related to the ArXiV paper [link]:

@misc{valtonenornhag-etal-2021-arxiv,
      title={Trust Your IMU: Consequences of Ignoring the IMU Drift},
      author={Marcus {Valtonen~Örnhag} and Patrik Persson and Mårten Wadenbäck and Kalle Åström and Anders Heyden},
      year={2021},
      eprint={2103.08286},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

Please cite the paper if you are using the code for (academic) publications.